ChatGPT Ads at ¥0.435 Per Impression: Should Businesses Heavily Invest in AI Traffic? A Practical Perspective from a Veteran Search Marketer

At the start of 2026, the marketing circle was abuzz with news: ChatGPT launched its ad beta program with a CPM quote of approximately $60 (around ¥435 RMB), making it the world's first large model platform to embed ads in generative results.

Many people's first reaction was: "Why is it so expensive?" Foreign trade practitioners began to feel anxious: "Is this the next traffic entry point? Should we heavily invest now?"

As someone with a decade of experience in search marketing – spanning website building, SEO, press release distribution, and now GEO projects – I've seen countless new channels that "look promising but are painful to invest in". So this article will only answer one core question:

Are large AI model ads worth heavy investment for enterprises right now?

The conclusion won't be extreme, but it will be as realistic as possible.

I. How Crazy is the AI Market? Tech Giants are Going All In

Before discussing AI advertising, let's look at a bigger reality: during the 2026 Spring Festival, almost all major domestic tech companies went all in on AI.

• ByteDance's DouBao appeared on the Spring Festival Gala stage

• Tencent poured over 1 billion RMB into promoting "YuanBao"

• Alibaba's Qianwen announced a 3 billion RMB ecosystem subsidy

• Baidu invested over 500 million RMB in promoting Wenxin Assistant

During the Spring Festival, AI was everywhere – on TV, in elevator ads, on WeChat Moments, and in splash screen ads. I even saw large-screen ads for Tencent YuanBao at Shanghai Hongqiao High-Speed Railway Station.

This is no longer a technical experiment, but a strategic bet. Capital is betting big, traffic is shifting, and user habits are changing.

This means AI has moved beyond the trend discussion phase to the stage of competing for traffic entry points. Against this backdrop, ChatGPT's ad trial is essentially not surprising.

Only one question remains: Is entering now an opportunity, or a trap?

II. Let's Do the Math: What Level is a $60 CPM?

Many people lack intuition about CPM, so let me put it in plain language.

CPM = Cost Per Mille = Cost per thousand impressions.

ChatGPT charges $60 for 1,000 impressions, equivalent to about 435 RMB, or approximately 0.435 RMB per impression on average.

Let's compare it horizontally with domestic and international advertising systems:

• Domestic: Douyin feed ads cost about 5–30 RMB CPM, Toutiao about 5–20 RMB, Xiaohongshu Juguang about 40–80 RMB, and WeChat Moments mostly 80–150 RMB.

• International: Google Search Ads CPM is about $10-20, Facebook/Instagram about $15-25.

ChatGPT's CPM is 3 times that of Facebook and 5-10 times that of traditional online ads, placing it directly in the "premium attention pricing tier".

At this point, the real question is not whether it's expensive, but whether it's worth the price.

III. Why Dares It Charge So Much?

From the platform's perspective, it's not selling mere impressions, but two scarcer things:

1. Audience targeting capabilities

2. Intensity of natural language intent

1. Current user quality is indeed relatively high

At this stage, the demographic that uses large models frequently has distinct characteristics: students, programmers, white-collar workers, product managers, creative professionals, and digital roles account for a higher proportion. This group generally has better education, stronger information acquisition abilities, higher acceptance of new tools, and greater ability to afford higher price points.

In other words, AI traffic currently comes with a layer of "cognitive screening". This is very similar to the early days of Zhihu and Bilibili.

However, it's important to note: this elite premium is most likely temporary. Internet history has repeatedly proven that as platforms scale up, user demographics inevitably become more diverse. AI will hardly be an exception.

2. What makes AI truly valuable is "active intent" (but it must be segmented by keywords)

Many articles like to summarize in one sentence: AI users ask questions proactively, so their intent is stronger. This statement is only half correct.

From my years of promotion experience, there's a hard and fast rule: not all questions indicate purchase intent.

• Weak-intent keywords: e.g., "how to build a website", "what is an ERP system", "how to set up an official website" – this type of traffic is usually "information-seeking", with high search volume, low click costs, and low conversion rates.

• Strong-intent keywords: e.g., "recommended website building companies", "which ERP system is best", "MES system quotation" – behind these keywords is often confirmed internal demand, supplier screening phase, and budget awareness. Truly valuable traffic is always this type.

The biggest uncertainty with AI advertising right now is: How precisely can the platform identify and match different levels of intent? If ads appear extensively alongside educational or tutorial-style questions, even with high-quality audiences, the final ROI may not be favorable.

IV. High-ticket B2B Industries Still Can't Do Without Baidu Paid Search in the Short Term

Let me be clear about this judgment:

Over the next 1-3 years, search ads will likely remain the main source of leads for high-ticket B2B industries.

This is especially true for high-ticket sectors such as ERP systems, MES systems, large industrial equipment, and production line equipment.

Because searching itself is a strong demand confirmation action. When someone searches for "recommended ERP systems" or "website building companies", it usually means internal demand has been confirmed, supplier screening has started, and a preliminary budget has been set.

Here are two real examples:

• A long-term client of mine: previously spent over 1.6 million RMB on Douyin in a year, getting over 2,000 "valid leads" but only closing 1 deal. Later, all budget was shifted back to Baidu paid search, as their product price starts at 300,000 RMB.

• A friend who runs a law firm: Baidu paid search had a ROI of about 1:4 (invest 10,000 RMB, sign 40,000 RMB in deals), but after switching to Douyin, there were many leads but almost no closed deals, losing over 30,000 RMB before giving up.

Why did they return to Baidu? Because they rely on search traffic, which delivers more accurate leads. This actually exposes many traffic illusions.

V. But AI is Quietly Changing "Pre-Purchase Mindset"

That said, while search remains stable for final conversions, AI is clearly capturing a portion of upstream research traffic – especially in the early stages of solution research, product selection, technical understanding, and comparative learning.

• Traditional path: Search → Read multiple sources → Summarize independently

• Current path: AI summary → In-depth search → Screening

This is why GEO has gained increasing attention from enterprises in recent years. Essentially, it's not replacing paid search, but competing for earlier cognitive entry points.

VI. AI Advertising Has Significant Noise

If you've done monitoring or research, you'll find two real issues:

1. A large number of peers, service providers, and industry practitioners frequently search for industry keywords for testing, monitoring, and demonstration purposes. This directly inflates impressions but may not bring real business opportunities.

2. Weak-intent questions actually account for a very high proportion in large models. There are numerous learning-oriented, research-oriented, and curiosity-driven questions. If the ad system fails to distinguish intent levels precisely, it's easy to see "impressive impressions and decent clicks but mediocre real conversions".

Feed ads actually went through this same phase in their early days.

VII. Is It Worth Heavy Investment Right Now?

If I have to give a very practical conclusion, my judgment is:

AI advertising is worth attention, but not yet worth going all in on.

To be more specific:

• Suitable for active testing: Enterprises with high price points, long decision cycles, content asset capabilities, and data analysis teams. Large companies with annual marketing budgets exceeding 10 million RMB can enter with a "small-budget forward-looking testing" approach.

• Not suitable for heavy investment now: Enterprises focused solely on performance, with cash flow pressure, low price points, or no content asset strategy at all.

Additionally, OpenAI has set a minimum spend commitment of $200,000 for ChatGPT ads – a threshold that's too high, suitable only for major players to test.

VIII. Practical Combination Recommendations for Enterprises

As of 2026, a relatively robust structure might be:

1. Search ads remain the main B2B conversion channel – high-priced equipment and enterprise software will be hard to replace in the short term.

2. Feed ads remain a scale reach tool – but material requirements are getting higher, and extensive targeting is becoming increasingly difficult.

3. Must start building GEO asset accumulation – this is the infrastructure for future AI recommendation weights, not just a nice-to-have.

4. Approach AI advertising with a testing mindset, not a gambler's mindset – small, iterative steps are more important than emotional heavy investment.

Truly rational enterprises never focus on a single channel, but build their long-term digital asset system simultaneously. No matter how traffic entry points shift, the official website remains the brand's "base camp", and search and AI recommendations still rely on content and trust.

In recent years, I've focused more on long-term initiatives: helping enterprises build solid website foundations, optimize SEO organic traffic, deploy GEO generative engine visibility, and continuously provide GEO training and press release services to assist enterprises in systematically improving their discoverability and credibility in search and AI environments.

If you're evaluating:

• Whether to upgrade/rebuild your official website

• Whether SEO is still worth investing in

• Whether now is the right entry point for GEO

• How to systematically layout brand exposure in the AI era

Feel free to reach out to discuss. Often, rather than blindly increasing budgets, it's more important to first get the structure right and lay a solid foundation. Traffic will change, but long-term compoundable capabilities are always timeless.